Fin

How does your team measure Fin AI performance?We're trying to understand how different teams think about this. When you review Fin's performance, what signals do you actually look at?- Resolution rate vs. target- Handoff rate- Reopened conversations- CSAT on Fin interactions- Something else entirely?And when Fin underperforms, how do you diagnose why — do you have a process for that, or is it mostly manual conversation review?Would love to hear how others are approaching this.

Hi everyone,I’m looking for some guidance on best practices for using Fin with a knowledge base that supports multiple products within the same platform.Our platform now has three core product areas, and we’ve structured our knowledge base to reflect that. Shared tools are documented as a single canonical “source of truth” article, with product specific context pages that explain how that tool behaves within each product and link back to the canonical guide.This hub and spoke structure works well for humans navigating the knowledge base, but I want to make sure Fin is also responding with the correct product context when a customer asks a question. The challenge is that some tools and terminology overlap across products, so if a customer asks about something like the email tool, the correct guidance depends on which product they are using. We want to avoid a situation where Fin surfaces guidance from the wrong product area simply because the t

Hi everyone 👋We’re currently exploring how to automate our carrier loss/damage claims workflow using Fin AI and would really value input from anyone who has implemented something similar.ContextWe manage parcel claims (e.g. Royal Mail and other carriers) on behalf of merchants. Today, this process involves significant manual review, validation, routing, and follow-ups. Our goal is to reduce manual workload while maintaining a strong audit trail and strict compliance with carrier policies.What we’d like to automate: Collect structured claim data (tracking number, order number, carrier, reason, invoice, proof of sale, damage photos, etc.) Validate required fields and supporting documents Dynamically enforce carrier-specific rules (claim windows, liability caps, required evidence) Route complete claims to the right internal team or BPO

We are currently attempting to test deployment of Fin for Zendesk Messaging and are encountering some issues where the Zendesk messaging widget is blank. We have tried multiple audience configurations with no luck.Why is the widget blank despite seemingly meeting the audience conditions and the deployment of Fin being live? Example of a conversation log in Fin of this:   When someone does not match the audience for Fin, shouldn’t they still experience the Zendesk Messenger/AI Bot or are they completely locked out of any messaging experience?

Our team hit 71% resolution rate with Fin last month. Looked great on the surface — but when we dug in, we had no idea which topics were dragging it down or what to actually fix.Intercom shows you the overall number but nothing underneath it. No topic breakdown, no trend over time, no way to know if a KB update actually moved the needle.Curious how other teams handle this. Are you exporting conversation data and analyzing manually? Using a third-party tool? Or just accepting the top-line number and moving on?(I've been building something to address exactly this — happy to share early access if there's interest.)

I want to be able to check if making a change to content will improve Fin’s answer to a question without first having to publish it. Is there anyway to do this? 

I just published a new Support Stack Solo episode showing how I built my first Fin Procedure inside Intercom.Instead of just answering questions, this procedure turns Fin into a qualification concierge for my website. It asks a few questions, stores attributes, scores the lead, and routes them to the right next step.If you're experimenting with Fin Procedures or AI workflows, this is a practical walkthrough of the full setup (including what broke during testing).Watch it here: Support Stack Solo E03: How to Build a Fin Procedure in Intercom (Real Qualification Concierge)

 I’d like to request assistance with turning off the image replies feature for our Fin AI Agent. Currently, Fin automatically includes up to three relevant images/GIFs from our support content in its responses, but we'd prefer to have text-only replies.I understand this isn't a self-serve setting and requires support assistance to disable. Could someone help us turn off this feature for our workspace?Thanks for your help! 

Fin Voice: Option to view what document or guidance the Fin Voice is using when answer the clients inquiry (same with email and chat). 

Ability to Rearrange Guidance. Make this section movable. This will help us Manage the old and new guidance better.  Put an alphabetical option Include created by date or by person for better visibility (aside from change log) but change logs disappears. Ability to rearrange by drag and drop so that we can make related guidance close enough for visual control.

Is there a way that FIN can be instructed to go to certain "Paths" within the workflow or it's just a dead-end when we let FIN handle it?  I noticed that when a user goes into the “Let FIN handle” node, there is no way out. The only option is to escalate/hand it over to a human agent. 

Hello - I want to connect FIN AI to my status page that is hosted on Atlassian. For this im using, I plan on using the Procedures feature and create a new connector.I need to make an API connection to a specific endpoint from Atlassian - does anyone has any recommendation on which endpoint I could use from Atlassian? Does anyone has the same setup. Any guidance is much appreciated. Thanks

 In Ep. 11 of Support Stack, Gabriela Passeri (Knowledge Manager at AutoDS) walks through how they made Fin actually understand their product.Their key move was to build internal, Fin-enabled glossaries mapping customer language to product terminology. Including synonyms, “avoid confusion with”, and even UI layout guidance.Super practical and very applicable if you have a complex B2B SaaS product with integrations!Full episode: I really hope it helps! 

Hi all! Long-time listener, first-time caller 🙃I’m Eamon, and I lead Customer Ops at Playground. We’re exploring expanding our support channels to include video calls (e.g. Zoom).We serve daycare providers, who are often short on time and not always deeply tech-proficient. We’re thinking video could be a more effective way to handle “How do I…?” questions - offering a more immersive, high-touch experience that helps them get unstuck quickly.I’d love to hear from anyone who has rolled out higher-touch channels like phone or video: What worked well? What didn’t? Anything you’d do differently? How did you position it alongside chat/email? If you’ve had success here, I’d really appreciate the chance to learn from you. Thanks in advance!

Hi All, We are attempting to use the Fin overlay on Zendesk Messaging. During several live deployment tests, we found that a quarter of our users are showing as ‘unrecognized’ which creates a duplicate profile in Zendesk. This wasn’t a problem when using Zendesk Messaging independently (no Fin overlay) We haven’t changed anything with our Zendesk Messaging widget authentication (JWT) Example code was reviewed and found to be functioning correctly per Fin and Zendesk SupportHas anyone else faced this issue and found a resolution? The duplicate profiles cause more work for agents to merge and impact our reporting if the profiles aren’t tied together. Thanks!

Increase Guidance Limit From 100 to 150 or more… Hello Team, we have reached 100 Fin guidance limit in total for Chat/Email and Voice. Can you separate this limiter instead? I mean...Like 100 for Voice and 100 for non-voice?Of can you just increase the limit to 150 if it’s easier to configure?  We are adding more features to as we go along the way. So, our other work space will surely reach the limit pretty soon. 

Fin is technically able to read dates, but in practice it often fails to reason about them correctly.ExampleFin currently sends out discount codes that are already expired, because it doesn’t reliably calculate whether a date is in the past, present, or future.Feature request    •    Fin should be able to compare dates to the current date    •    Fin should not mention or suggest expired offers    •    Ideally, Fin should be aware of:    •    “valid until”    •    “expired”    •    “not yet active”Why this matters    •    Prevents customer frustration    •    Avoids incorrect promises (e.g. expired discounts)    •    Increases trust in AI-generated responses 

Hey again! I noticed there is an option to collect Fin’s attributes data which is great, but is there any chance to be able to collect attributes’ data without Fin answering? Is this something that we can expect any time in the future?   

Using the FIN workflow, I’d like to trigger a bot survey using a URL. I’d also want some metadata passed over from FIN through the URL to our survey tool. How can I accomplish this?  Metadata examples:-FIN Conversation ID-Category

Hi All, What are the benchmarks for bot-resolved CSAT? Our team is looking for a baseline as we implement our first round of AI/Fin improvements. Thanks! 

I’m looking for best practices on how concrete service companies such as AMC Concrete can manage customer communication more effectively during residential and commercial projects.What tools or workflows work best for handling client inquiries, project schedules, progress updates, pricing clarity, and post-project follow-ups without putting extra pressure on the team?

Our company offers software which contains databases our end users need for their work. The databases sometimes use difficult terminology that is hard to interpret for our users. In an attempt to improve the ‘readability’ of the databases, I have recently uploaded a bunch of ‘metadata’ associated with the most used databases in our software - breaking FIN as a result.Prior to starting the upload December 2025, I had contact multiple times with my Intercom account manager to determine the ‘optimal’ way to upload a large amount of qualitative (text) data to Intercom. The data on our end originally is stored in JSON format - a format we can’t directly upload directly to Intercom. After some discussions, we opted to split the JSON file into smaller .txt files, which we manually uploaded as ‘snippets’ to Intercom. We have spent a considerable amount of time setting this up. Now the system is ‘stuck’ since Christmas 2025, with no viable solution yet in place. No new content we uplo

How do you all tie your CSAT reporting together? We use Fin along with Zendesk and house data externally in Domo. It doesn’t appear possible to send Fin CSAT over in the Zendesk ticket that is created for tracking, and it looks like our only option is manually exporting CSAT CSV’s and then importing that into Domo. Is this correct?Are there any better solutions?

I'm looking for a streamlined QA process for managing Fin’s performance as a one-woman team. Would love some suggestions for best practices for maintaining quality without getting overwhelmed with all the reports and data Intercom provides for Fin performance.

It would be extremely helpful if agents could give Fin instructions directly inside an ongoing conversation, without having to update global content or wait for retraining.ExampleAn agent could write internal instructions like:    •    updated information    •    temporary notes    •    clarifications    •    corrections…and Fin would immediately apply these instructions and continue the conversation with the customer using the updated information.Why this matters    •    Allows quick corrections when Fin is about to provide wrong or outdated information    •    Enables real-time steering of AI responses without agent replies    •    Especially useful for temporary changes (e.g. delivery delays, special cases, exceptions)This would make Fin far more flexible and safer to use in live support situations.